NVIDIA Is Looking Into Reports Of Burnt RTX 4090 12VHPWR Adapters As More Pics Emerge
We reported on this issue on Monday, where a Reddit user posted pictures (including the above) of his melted power connector. Several more pictures of this phenomenon have emerged, but notably, all of them have been on GeForce RTX 4090 cards using the official adapter—and none of them have appeared especially bent or abused.
Source: PCI-SIG via Seasonic
For clarification, we should explain. Before the GeForce RTX 4090 came out, the PCI-SIG, which has ratified the 12VHPWR connector as a standard plug, put out a document explaining that applying cable tension to the 12VHPWR connector, or using it for more than 40 plug-unplug cycles, could cause the connector to fail and overheat. Naturally, some people assumed that the 12VHPWR connector itself was at fault, and perhaps a poor design.
The thing about that is that the connector itself isn't completely new. The 12-pin power connectors on the GeForce RTX 3070 and higher Ampere cards is mechanically the same connector, and we haven't heard about any failures of this sort on those cards. You could say that it's the higher power draw of the RTX 4090 causing this problem, except it's actually very efficient and uses less power than the RTX 3090 Ti while gaming.
Image from GALAX showing hot wires when improperly plugged. Source: Twitter
Hassan Mujtaba on Twitter posted some images that apparently originate with graphics card manufacturer GALAX of that company's testing with the 12VHPWR connector. Unsurprisingly, a poor connection (such as from insufficient insertion force) can cause the connector to get very hot very quickly. This is due to uneven transfer of current across the pins, and was an issue mentioned by PCI-SIG, but given the wide clip on the outside of the connector, we find it unlikely that enthusiasts are limp-wristing the 12VHPWR plugs on their GPUs.
None of the connector failures that have been observed by end-users have been the fault of a connector exceeding its plug-unplug cycles, and most of the failures seem to have been on cards with minimal and expected bending of the wires. So saying, it does look ultimately as if the adapters themselves are at fault for the failures.
Curiously, several sources have reported that NVIDIA is demanding burned cards from users and AIBs, which is something that it has never done before, even during the "Space Invaders" issues with the RTX 20 series. Kudos to the company for investigating the failures, but it seems like it should probably be investigating its partnership with the adapter manufacturer, not AIBs' card designs.
